Can a character be smarter than its author? Sherlock Holmes is surely smarter than Arthur Conan Doyle, but how do you do that? How can a novelist feign an intelligence that he doesn’t have? Goethe said that courage and modesty are the only two true virtues because there is no way to simulate them. Does the same thing happen with intelligence, that there is no way to pretend it? Now that half the planet has been abducted by ChatGPT, the fashionable digital chatterbox, we have a new angle from which to approach all those annoying questions.
To begin with, there is no doubt that ChatGPT has passed the Turing test. The great British mathematician Alan Turing, a pioneer of computing and artificial intelligence, proposed in 1950 that a machine would be intelligent when it could impersonate a person in blind conversation. The texts produced by ChatGPT may seem more or less accurate, pragmatic or fanciful, but it is frankly difficult to distinguish them from what a person would say. Homo sapiens instead. If you ask him why there is something rather than nothing—an unanswered question—he begins by acknowledging that this is a very deep question and then gives you half a veronica and ends by answering something else. It is exactly what a person would do, and the Turing test is therefore passed.
But is that intelligence? Or just fake it? We know by now that ChatGPT composes children’s stories and academic articles, tells jokes, writes computer code and passes the selectivity. The clinical psychologist Eka Roivainen has put poor ChatGPT to an IQ test (intelligence quotient) as he does to his patients in a Finnish hospital. He has used the most common test (Wechsler, or WAIS), where the human average is 100, 10% of people reach 120, and the smartest 1% reach 133. ChatGPT has scored 155. Only 0, 1% of humans can be compared to him. That is 50,000 people in Spain, 3,000 in Madrid, 200 in Carabanchel and none in my house, rounding off a bit. Flat.
I know what you’re thinking: IQ doesn’t measure intelligence, right? It is what we humans always do in our undeclared war against the machines. If Deep Blue beats Kasparov, we conclude that chess is not a sign of intelligence. If a machine passes the Turing test, we don’t throw away our pride, but the Turing test. If ChatGPT has an IQ to join the Mensa club, it will be that IQ does not measure intelligence, but something else that we don’t know what it is. Anything but admit that a machine surpasses us in some intellectual activity.
That said, it’s true that many computer scientists have thought for years that the Turing test, IQ tests, and other black-box approaches—where you just see the results without examining what’s going on in the mind of the machine— They fall far short of the magnitude of this problem. If you ask him what is the first name of the father of Sebastián’s children?, ChatGPT comes up for peteneras, arguing that it lacks context. What a smartass
Subscribe to continue reading
Read without limits